Seasonal Water Mass Analysis for the Straits of
Juan de Fuca and Georgia
Diane Masson*
Institute of Ocean Sciences
P.O. Box 6000
Sidney BC V8L 4B2
[Original manuscript submitted 7 September 2004; in revised form 6 January 2005]
A quantitative analysis of water masses in the coastal waters of southern British Columbia is performed with the Optimum Multiparameter (OMP) analysis method that optimizes the use of a hydrographic
dataset by solving an over-determined linear set of mixing equations. The method is applied to a seasonal dataset
collected over five years in the Strait of Georgia, a large semi-enclosed coastal basin, as well as in Juan de Fuca
Strait, its main connection to the Pacific Ocean. Abundant freshwater discharge into the coastal basin forces an
estuarine exchange with oceanic shelf water. Six water characteristics of five source water types are used to
obtain mixing proportions over the estuary for each of the four seasons. The model results are found to corroborate known aspects of the local dynamics such as the presence of a deep shelf inflow into Juan de Fuca Strait and
of the Columbia River plume in winter at the mouth of the strait. The analysis also quantifies lesser known features of the region, such as the characteristics of the mid-depth intrusions within the Strait of Georgia and the
marked effect of remineralization on nutrient distributions in the deep water of the coastal basin.
ABSTRACT
RESUMÉ [Traduit par la rédaction] On effectue une analyse quantitative des masses d’eau dans la zone maritime
littorale du sud de la Colombie-Britannique par la méthode de l’analyse multiparamétrique optimale (AMO) qui
optimise l’utilisation d’un ensemble de données hydrographiques en résolvant un ensemble linéaire surdéterminé
d’équations de mélange. La méthode est appliquée à un ensemble de données saisonnières recueillies sur une
période de cinq ans dans le détroit de Georgie, un vaste bassin côtier semi-fermé, ainsi que dans le détroit de
Juan de Fuca, son lien principal avec le Pacifique. L’arrivée d’une grande quantité d’eau douce dans le bassin
côtier force un échange estuarien avec l’eau océanique de la plate-forme. On utilise six caractéristiques
hydriques de cinq types d’eau sources pour obtenir les proportions de mélange dans l’estuaire pour chacune des
quatre saisons. On trouve que les résultats du modèle s’accordent avec les caractéristiques connues de la
dynamique locale, comme la présence d’un courant profond entrant dans le détroit de Juan de Fuca depuis la
plate-forme et la présence du panache du fleuve Columbia en hiver à l’embouchure du détroit. L’analyse quantifie
aussi des aspects moins connus de la région, comme les caractéristiques des intrusions à profondeur moyenne
dans le détroit de Georgie et les effets prononcés de reminéralisation sur la distribution des nutriants dans les
eaux profondes du bassin côtier.
1 Introduction
The Straits of Georgia and Juan de Fuca are the two main
bodies of water separating Vancouver Island and the mainland of the southern coast of British Columbia (Fig. 1). The
Strait of Georgia is a semi-enclosed coastal basin that extends
about 220 km along the coast, with water depths up to 420 m
within its central deep basin. Its main connection with the
open ocean is through Juan de Fuca Strait to the south, which
is also the outlet for the Puget Sound Basin. Narrow channels
between a series of small islands, as well as shallow sills, constrict the flow between the two straits, with most of the water
flowing through the deeper western channel comprised of
Boundary Pass and Haro Strait. To the west, Juan de Fuca
Strait opens up into a more regular channel-like basin of
about 20-km width, reaching water depths of 250 m at its
mouth. At its northern end, the Strait of Georgia is connected
to the open ocean through constricted and shallow channels,
with cross-sectional areas much smaller (7%) than those of
the southern route (e.g., Waldichuk, 1957).
The three coastal basins form a large estuary in which a complex circulation is shaped by the basin morphology, a large and
variable freshwater runoff, strong tidal currents and orographically steered coastal winds. Early descriptions of the physical
oceanography of the Strait of Georgia and Juan de Fuca Strait
were given by Waldichuk (1957) and Herlinveaux and Tully
*Corresponding author’s e-mail: [email protected]
ATMOSPHERE-OCEAN 44 (1) 2006, 1–15
© Canadian Meteorological and Oceanographic Society
2 / Diane Masson
Fig.1
Study area and locations of the sampling station (small circles). The shaded line indicates the position of the along-strait section.
(1961), respectively. Later reviews of the region have been
completed by Thomson (1981) and LeBlond (1983).
Rivers and small creeks discharge a significant amount of
fresh water into these basins. The most important source of
fresh water is the Fraser River. It has pronounced seasonal
flow variations with low winter discharge rates and much
larger flow during spring freshet (Fig. 2). Tides in the area are
mixed, mainly semidiurnal, and strong tidal currents are
found in the constricted and shallow passages, producing significant mixing of the water column.
During winter months, strong south-easterly winds predominate while during summer weaker north-westerly winds
are common. Associated with the seasonal change in domi-
nant wind direction is a transition from downwelling (winter)
to upwelling (summer) conditions over the continental shelf
(e.g., Thomson, 1981). As a consequence, the deep estuarine
return flow is colder and saltier (denser) during summer.
These changes in both freshwater runoff and atmospheric
forcing significantly modulate the estuarine circulation of the
coastal basin on the seasonal timescale.
In this paper, the Optimum Multiparameter (OMP) analysis is used to estimate the mixture of source water types that
best describes the composition of the local water masses. The
method has been previously applied to various oceanic
regions such as the continental margin off Vancouver Island
(Mackas et al., 1987), and the thermocline in the Indian
Seasonal Water Mass Analysis for the Straits of Juan de Fuca and Georgia / 3
Fig. 2
Mean daily Fraser river discharge measured at Hope, BC, (130 km upstream) and timing of sampling surveys. The mean discharge values were computed
over the period 1912–2003.
Ocean (You and Tomczak, 1993). The analysis is performed
on a seasonal basis to examine changes in water mass composition during the year. A similar seasonal water mass study
for the western South Atlantic Ocean was conducted by
Maamaatuaiahutapu et al. (1994) and for the thermocline in
the Indian Ocean by You (1997).
In the next section, the dataset used in the analysis is
described. A brief description of the method is given in
Section 3. The results are described in Section 4, followed by
some concluding remarks.
2 Data
In 1999, a field program was established to sample the estuary a few times each year in an attempt to gain a better under-
standing of the regional circulation and of its seasonal variability. During each survey, CTD profiles and water samples
were collected at twenty stations lying along the main axis of
the estuary extending from the mouth of Juan de Fuca Strait
up to the northern end of the Strait of Georgia (Fig. 1). In
addition to temperature and salinity, nutrient salts (nitrate,
phosphate and silicate) as well as dissolved oxygen were
measured at these stations, near the surface and bottom as
well as at depths of 5, 10, 20, 30, 40, 50, 75, 100, 125, 150,
175, 200, 250, 300, 350, and 400 m.
A total of nineteen surveys were conducted over the five-year
period from 1999–2003 (Fig. 2). Sampling was carried
out seasonally during periods of: (1) early freshet (April),
(2) peak freshet (June/July), (3) end of freshet period
4 / Diane Masson
(August/September), and (4) low winter discharge (January/
December). For the analysis, the dataset has been synoptically
divided according to these four seasons (referred to as spring,
summer, fall, and winter) and the water mass analysis applied to
the resulting four datasets.
3 Method
A linear inverse mixing model, the OMP analysis, is applied
to hydrographic, nutrient, and dissolved oxygen data measured seasonally in southern British Columbia (BC) coastal
waters. At the location of each data point, and for each season, the model estimates the mixing fractions of predefined
source water types. For each data point, the contributions
from all water sources are obtained by finding the optimal linear mixing combination in the six-parameter space defined by
temperature, salinity, oxygen and nutrients which minimizes
the residuals in a non-negative least squares sense. The mass
conservation condition adds an additional mixing constraint
to the minimization problem.
The problem can be expressed mathematically as the minimization of
D2 = ( Ax − b)T W −1 ( Ax − b),
(1)
where A is a matrix containing the tracer values of the source
water types, b is a vector containing the observed parameter
values for a given sample, x is a vector containing the mixing
fractions, and D is a vector containing the residuals. The
matrix W contains the weights ascribed to the various parameters to reflect differences in measurement accuracy and
environmental variability. In addition, the solution involves
two constraints: all source contributions are to be positive,
xi ≥ 0,
(2)
and the fractional contributions from all sources must add up
to near unity,
∑ xi
= 1 + ε,
ε << 1.
(3)
Before solving the optimization problem, both elements
of the matrix A and the vector b in Eq. (1) are normalized
to make parameters of incommensurable units comparable,
as in
Aij′ =
( Aij − Aj ) , and b′ = (b j − Aj ) .
σj
j
σj
(4)
Here, Aj is the seasonal mean of parameter j for all source
water types i, and σj the associated standard deviation.
The elements of the weight matrix, W, can de determined
by various procedures. Mackas et al. (1987) obtained their
weights from the sum of the sample error estimates plus the
covariances between tracers of the source water types.
Tomczak and Large (1989) used a diagonal weight matrix
based on the parameter variability in the source regions. Here,
we follow closely the latter, and compute
Wj =
σ 2j
δj
,
(5)
where σj measures the ability of parameter j to resolve differences between water sources and δj is the mean of the water
type variances for parameter j. The constrained optimization
problem is solved by the Generalized Reduced Gradient
Method (e.g., Lasdon et al., 1978). Finally, the model results
obtained at each measurement point are then interpolated and
contoured on the along-strait section of Fig. 1.
4 Results
a Mean Tracer Distributions
As for most temperate estuaries, the local water density is mainly controlled by salinity, with temperature acting largely as a
passive tracer. The seasonal cycle of salinity in the study area
has been recently described and modelled by Masson and
Cummins (2004) using data from the same seasonal surveys.
Water properties of the coastal basin have been studied in various investigations, such as the early work of Waldichuk (1957),
and reviewed by LeBlond (1983). Some aspects of the nutrient
cycles and distributions within the estuary have been described
by, among others, Tully and Dodimead (1957), Lewis (1978),
and Harrison et al. (1991). However, because the present dataset
is quite extensive, covering the entire coastal basin over five
years, it allows for a description of unprecedented detail of the
local climatology for the various water properties.
The five-year mean distributions along the main axis
(shaded line in Fig. 1) of the estuary for each of the tracers
(temperature, nitrate, phosphate, silicate, and dissolved oxygen) are presented in Fig. 3. The figure also shows density
contours, which are indistinguishable from salinity contours,
to help delineate known features of the local dynamics. The
estuarine circulation draws a mass of dense, cold, nutrientrich water with low dissolved oxygen at depth into Juan de
Fuca Strait from the continental shelf (e.g., Mackas et al.,
1987). The mean distribution also shows a fresh and relatively warm surface river plume extending over most of the
surface of the Strait of Georgia, with reduced nutrient and
high oxygen concentrations associated with high primary production (e.g., Harrison et al., 1991). At both ends of the Strait
of Georgia, strong tidal currents are quite effective in mixing
the water column and in reducing the local stratification (e.g.,
Parsons et al., 1981). The deep water mass in the Strait of
Georgia, with high nutrient and low oxygen concentrations
(e.g., Tully and Dodimead, 1957), is partially isolated from
the active estuarine dynamics by the presence of shallow sills.
Vigorous mixing in these shallow areas controls water
exchange within the estuary and tends to favour recirculation
of mixed water in the inner deep strait, limiting direct access
of higher salinity shelf water. Hence, vertical gradients of
density and temperature in the deep strait are relatively weak.
Seasonal Water Mass Analysis for the Straits of Juan de Fuca and Georgia / 5
TABLE 1
Property values for each of the source water types for the four seasons given clockwise from the top left corner as spring, summer, fall, winter.
Source water type
SW1: river
SW2: deep south
SW3: pre-season
SW4: surface south
SW5 : deep north
Temp (°C)
Salinity
Oxygen (mL/L)
Nitrate (µM)
Phosphate (µM)
Silicate (µM)
9.1
8.1
6.8
8.0
9.4
9.2
12.5
13.9
6.4
6.8
9.1
8.9
27.3
27.5
33.9
33.2
31.2
31.1
25.1
26.2
33.9
33.9
31.0
30.9
7.5
6.1
2.1
3.1
2.9
3.2
6.1
6.0
1.8
1.6
3.0
3.5
10.1
26.3
33.4
28.1
28.8
28.4
11.1
6.8
35.5
34.8
29.4
28.1
1.1
2.3
2.6
2.4
2.8
2.6
1.2
0.9
2.7
2.7
2.8
2.6
25.0
53.5
51.5
43.3
58.1
54.0
33.3
30.0
57.3
56.1
61.4
56.1
9.7
8.9
9.0
8.7
9.4
30.3
30.4
30.3
30.5
30.3
6.5
3.9
3.7
3.7
3.5
7.3
28.4
28.4
29.2
26.0
1.0
2.6
2.6
2.7
2.6
17.8
56.8
56.2
59.8
54.8
In addition to highlighting known characteristics of the
estuary, Fig. 3 also reveals interesting features that have
received little attention in the past. For example, the estuarine
return flow entering the Strait of Georgia from the south may
be discerned as a tongue of relatively low dissolved nutrient
and high oxygen concentrations around sill depth. These subsurface intrusions form in the southern strait where vigorous
tidal currents flowing through a maze of narrow passages
cause partial recirculation of the fresh surface outflow. In the
five-year mean fields, the mid-water intrusions are centred at
100-m depth, along the 23.5 density contour, and extend
northward across most of the Strait of Georgia (Figs 3b–3e).
The water properties of these intrusions do vary during the
year (e.g., Masson, 2002) and this aspect of the circulation
will be examined more closely in the next section.
Few datasets have been published on phosphate and silicate
concentrations in the deep Strait of Georgia. In an early study,
Tully and Dodimead (1957) examined samples taken in the
upper 150 m only and found high levels of these nutrient salts
below the euphotic zone. They attributed the high concentrations of phosphate to the shelf water entering the estuary at
depth and the high silicate levels to the Fraser River waters
known to be rich in silicate (e.g., Harrison et al., 1991). Figure
3 shows high concentrations of dissolved phosphate (2.75 µM)
and silicate (60 µM) in the deep Strait of Georgia. However,
these concentrations are clearly higher than in the dense inflow
entering from the shelf. The analysis below allows a closer look
at the variability of nutrient concentrations at depth within the
basin as well as a discussion on the probable sources such as
remineralization within the deep basin.
b Determination of Source Water Types
The first step in the local application of the OMP analysis is
to define the source water types for the study area. The fact
that the estuary is a semi-enclosed basin simplifies the problem somewhat relative to typical open ocean applications of
the method. The local bathymetry and the estuarine nature of
the coastal basin help define the first two water sources for
the estuary: surface freshwater inflow in the Fraser River
plume (SW1), as well as deep inflow from the shelf at the
mouth of Juan de Fuca Strait (SW2). A third “virtual” water
type is added to the analysis in order to represent water resident in the deep Strait of Georgia which can be, at times, stagnant (SW3). The latter does not correspond to the typical
definition of a source water type where measurement points
are considered to be downstream of a number of known water
sources. However, it is necessary to represent the contribution
to the water masses present in a given season from the water
residing in the deep basin during the preceding season. This
is consistent with a known flushing time for the basin of about
a year (e.g., Li et al., 1999).
A fourth source water type, SW4, is added for the wintertime analysis to take into account the surface water intrusion
into Juan de Fuca Strait of fresh warm water originating from
the Columbia River. These surface intrusions are known to
enter the strait regularly during winter storms (e.g., Hickey
and Banas, 2003), bringing relatively warm, low salinity and
low nutrient water northward along the coast. The signature
of this water source is clearly identified in the mean distributions (Fig. 3) as a surface lens of warm, fresh and low nutrient water near the southern boundary. The inclusion of the
latter source for the winter analysis significantly improves
the performance of the mixing model. Finally, in a complementary analysis performed to look at the potential importance of intrusions from the north, a fifth source is identified
as the subsurface estuarine inflow through Johnstone Strait
(SW5).
Tracer characteristics for each water source are obtained, for
all four seasons, by averaging tracer values measured near the
point of origin for that source over the five years of data collection. More specifically, the characteristics of the water sources
are obtained from data measured at: stations 39 and 42 above
15 m for SW1, stations 75 and 102 below 200 m for SW2, stations 27 and 39 below 200 m for SW3, stations 75 and 102
above 20 m for SW4, and at station 16 below 100 m for SW5
(see Fig. 1 for station locations). Property values for each of the
source water types used in the analysis are given in Table 1, with
the four seasons given clockwise from the top left corner as
spring, summer, fall, winter.
c Mean Source Water Type Concentrations
In the first application of the mixing model, all parameters
are considered conservative. This is a reasonable first-order
assumption considering that, in this very dynamic coastal
basin, the effects of advection and turbulent diffusion are generally expected to outweigh local forcing and biogeochemical
effects. However, in some of the domain, such as near the surface, this assumption is invalid and may result in large
6 / Diane Masson
Fig. 3
Five-year mean distributions along the main axis of the strait for (a) temperature, (b) nitrate, (c) phosphate, (d) silicate, and (e) dissolved oxygen.
Numbers along the top axis indicate the location of the stations identified in Fig. 1.
Seasonal Water Mass Analysis for the Straits of Juan de Fuca and Georgia / 7
Fig. 4
Mean contributions for the source water type (a) SW1, (b) SW2, (c) SW3, (d) SW4 and (e) residuals obtained from the seasonal mixing analysis.
8 / Diane Masson
residuals, indicating a poor fit to the data. In the next section,
an attempt will be made to account for biogeochemical
processes by introducing a new column to the water type
matrix which represents these changes.
At each measurement location, the analysis is first performed with six parameters and the first three (four in winter)
water sources. The resulting mixing coefficients obtained by
averaging the results of all seasons are given in Fig. 4, along
with residuals (D2 of Eq. (1)). Many aspects of the distributions so obtained can be associated with known features of the
local estuarine circulation, lending credibility to the mixing
model. In particular, the dense shelf water inflow (SW2) and
the fresh surface river plume (SW1) mix in the area of strong
tidal currents located between Victoria Sill and Boundary
Pass. The resulting mixed water then advects out of Juan de
Fuca Strait as a surface outflow, and into the Strait of Georgia
as subsurface intrusions. In the deep Strait of Georgia, below
sill depth, water can at times be stagnant as indicated by the
high fraction of SW3, particularly evident in the northern half
of the basin.
The winter analysis was first performed without the inclusion of SW4. But, in this case, the mixing model could not
achieve a reasonable solution, giving unacceptably high residual values near the surface at the mouth of Juan de Fuca Strait.
The winter data were then analysed including an additional
source water type (SW4) representing the surface water influence of the Columbia River plume. This reduced the residuals
in this area of the domain considerably and indicated high concentrations of SW4 at the surface, near the mouth of Juan de
Fuca Strait. This water mass does not appear to be present during the rest of the year and, accordingly, the inclusion of SW4
did not significantly improve the solution for the other three
seasons. This is an illustration of how residuals provide a way
of assessing the adequacy of the source water type definitions.
Over most of the area, residuals are relatively low for all seasons (Fig. 4e). However, one clear exception occurs near the
surface of the Strait of Georgia, within the river plume. This is
to be expected given that biological processes and direct heat
exchange with the atmosphere are more intense within the
more stratified surface plume. But, to a lesser extent, larger
residuals are also obtained near the bottom of the deep basin of
the Strait of Georgia, indicating the possibility of non-conservative properties. In this case, remineralization and possibly denitrification is expected to affect local nutrient and dissolved
oxygen concentrations in the deep basin. Finally, larger residuals in the SW4 area are due to the intermittent nature of the
presence of the Columbia River plume in winter at the surface
of the mouth of Juan de Fuca Strait.
d Seasonal Changes of Mixing Fractions
Seasonal changes in water mass composition are presented in
Figs 5 to 7 for each of the three main source water types. In
Fig. 5, the analysis shows that, as expected, the water associated with the Fraser River plume, SW1, follows the seasonal
cycle of the river discharge closely (see Fig. 2) with maximum extent occuring in early summer during freshet. The
subsurface recirculation of the fresh water within the Strait of
Georgia is also stronger in early summer when the estuarine
circulation is at its maximum strength (Fig. 5b). In winter, a
less stratified and thicker plume forms as a consequence of
increased turbulent mixing during winter storms (Fig. 5d).
The seasonal contribution from the deep estuarine return
flow entering Juan de Fuca Strait, SW2, is contoured in Fig.
6. The penetration of the dense shelf water is maximal in summer when freshwater discharge into the estuary is high and
upwelling is prevalent on the coast. This source water type
contributes significantly to the mid-depth intrusions penetrating the Strait of Georgia from the south, constituting nearly
50% of the incoming subsurface water mass. The depth of
these intrusions varies with season, being restricted to sill
depth (around 100 m) in spring and reaching the bottom in
summer and fall, during periods of deep water renewal (e.g.,
Masson, 2002). Finally, in winter, when the estuarine circulation is at its weakest, there is little evidence of these subsurface intrusions entering the Strait of Georgia.
Figure 7 tracks the changes with season of the contribution
of the water mass residing in the deep Strait of Georgia during the previous season, SW3. It indicates that, in winter and
spring, the water of the Strait of Georgia below the sill depth
(about 100 m) is stagnant, with a large proportion of pre-season water. However, in summer and fall, the analysis indicates that dense intrusions from the south enter the deep basin
and displace the ambient deep water mass. This is consistent
with the known seasonal cycle of the deep water properties
which indicate that the main water renewal events occur in
early summer and in the fall (e.g., LeBlond et al., 1991). It is
also interesting to note that, for most of the year, sill depth
intrusions create, within the Strait of Georgia, a three-layer
structure below the surface plume: new intruding water at sill
depth becomes intertwined with pre-season deep water. The
intruding intermediate layer is characterized by relatively
high dissolved oxygen and low nutrient concentrations as
well as cold (warm) temperature in the spring (fall), creating
subsurface maxima and minima in the vertical profiles of
these tracers (see Fig. 3).
e Sensitivity Analysis
To verify the robustness of the results relative to uncertainties
in the source water type definitions, the analysis was repeated 100 times, varying randomly the seasonal mean characteristics of each of the source water types, Aij, within plus or
minus one standard deviation of all measurements contributing to the mean. The average of the results from these 100
simulations was then computed and found to be practically
identical to the original results.
In the initial analysis, assumptions were made regarding
which water sources contribute significantly to the coastal
water mass. Accordingly, four sources were used: the river
plume, the deep inflow from the south, the pre-season deep
water, as well as surface winter intrusions from the south. This
first model, in which possible intrusions from the north were
not accounted for, appears to perform generally well, including
Seasonal Water Mass Analysis for the Straits of Juan de Fuca and Georgia / 9
Fig. 5
Contributions from the source water type associated with the Fraser River plume, SW1, for all four seasons.
10 / Diane Masson
Fig. 6
Contributions from the deep southern inflow, SW2, for all four seasons.
Seasonal Water Mass Analysis for the Straits of Juan de Fuca and Georgia / 11
Fig. 7
Contributions from the pre-season water source, SW3, for all four seasons.
12 / Diane Masson
Fig. 8
Contributions from subsurface intrusions from the north, SW5, averaged over all seasons.
in the northern end of the domain (Fig. 4e). Intrusions from the
northern end of the strait have been mentioned in the past (e.g.,
Waldichuk, 1957; LeBlond, 1983) but have rarely been clearly
identified. To examine the impact of such intrusions from the
north on the model results, the basic model is now modified by
adding a fifth source representing subsurface water originating
at the northern end of the Strait of Georgia (SW5 of Table 1).
Figure 8 gives the resulting SW5 contributions averaged over
the four seasons. Contributions from other sources are generally unchanged except for SW3, the deep pre-season type, which
is reduced in the area of high SW5 values. The contours indicate a distribution of SW5 consistent with southward advection
and mixing of subsurface water from the northern channels at
about the sill depth, 80 m. The intrusions are of lesser extent
and shallower than the intrusions from the south, owing to the
relatively small cross-section of the northern channels, a shallower sill depth and the lower density of the intruding water.
The depth of penetration appears to change slightly with season, being shallowest in late summer when the new water type
is distributed right under the river plume.
It is legitimate to ask whether or not the addition of the fifth
source improves the performance of the mixing model. In
order to compare the goodness of fit of the two applications,
the sum of residuals was computed in both cases. However,
because the variables in Eq. (1) are normalized and the normalization changes with source water type definitions (Eq.
(4)), we compute, for each sampling point, a new residual
parameter, independent of the normalization:
D̂2 = ( Ax − b)T ( Ax − b)
(6)
Figure 9 presents the value of D̂2 as a function of depth, averaged over the Strait of Georgia, for the analysis with and with-
out SW5. It shows that adding the fifth water source markedly
improves the results around the depth at which the northern
intrusions enter the strait. Obviously, the addition of a source
water type defined within the domain increases the degrees of
freedom of the mixing model and, consequently, locally
reduces the residuals. Therefore, although the present results
are consistent with the presence of such intrusions, the model is
not capable of unequivocally determining their existence.
As noted earlier, the conservation of water properties is an
assumption inherent in the mixing model. This is bound to fail
in regions where significant biogeochemical changes occur,
such as near the ocean surface. A modified version of the
method, the so-called extended OMP, combines mixing of
predefined source water types with time-integrated biogeochemical changes in oxygen and nutrients due to remineralization (e.g., Karstensen and Tomczak, 1998). The method
consists of including an additional column in matrix A of Eq.
(1) to represent the increase in nutrients and reduction in dissolved oxygen through the water column due to remineralization. Accordingly, a new unknown, the amount of
biogeochemical change in phosphate concentration (∆P), is
added to the problem and estimated by the optimization technique at each sampling location. The extended OMP method
is applied to the present dataset but, in contrast to previous
applications, we allow here changes in phosphate concentrations (∆P) to be either positive or negative, in an attempt to
account for remineralization as well as primary production.
Associated changes in nitrate, and dissolved oxygen are then
obtained from the estimated change in phosphate by using the
classical Redfield ratios of [16:-138] for [∆Ν/∆P: ∆Ο2/∆P]
(Redfield et al., 1963).
To complete the model formulation, a ∆Si/∆P ratio is also
needed. In contrast to nitrogen and phosphorus, silicon is not
part of the organic soft tissue but is an important constituent
Seasonal Water Mass Analysis for the Straits of Juan de Fuca and Georgia / 13
Fig. 9
Residuals, D̂2, as a function of depth , averaged over the Strait of Georgia, for the basic analysis (full line), for the case including a fifth source water
type (dotted line), and for the extended OMP case (dashed line).
of silica-shelled phytoplankton such as diatoms. It is transported down the water column by the sinking of these organisms. Because the ratio ∆Si/∆P varies with species and
because the proportion of diatoms to other phytoplankton
varies with location (e.g., Redfield et al., 1963), the ratio of
dissolved silicate to phosphate also changes significantly. By
using a best fit to the measured phosphate-silicate concentrations, we estimated the local ∆Si/∆P remineralization ratio to
14 / Diane Masson
Fig. 10
Mean value for the biogeochemical changes in dissolved phosphate, ∆P, estimated by the extended OMP method.
be 23. This value is consistent with the estimates of Hupe and
Karstensen (2000) for the Arabian Sea who found an average
ratio of 21 in the upper 2000 m of the water column.
The basic mixing model, with the first four source water
types only, is then applied, allowing biogeochemical changes
in the oxygen and nutrient concentrations. Overall, mixing
fractions of the different source water types are similar to the
basic model results. Figure 10 shows the contours of the biogeochemical changes in phosphate, ∆P, estimated by the
extended OMP, averaged over the year. As expected, large
negative values for ∆P at the surface of the Strait of Georgia
indicate a high level of photosynthesis in the euphotic zone.
Surface values are largest in April, during the spring bloom.
Also, in summer, there are indications of primary production
in the surface waters at the mouth of Juan de Fuca Strait. On
the other hand, remineralization is evident in the deep Strait
of Georgia where ∆P takes larger positive values. This is in
agreement with high levels of dissolved nutrient and low dissolved oxygen values measured in the deep Strait of Georgia
(see Figs 3b–3e). For the case of silicate, such a high concentration likely contributes to the presence of rare siliceous
sponge reefs which have been found at depth in the Strait of
Georgia (Conway et al., 2004). The residual parameter D*
(Eq. 6) obtained with the extended OMP is compared in Fig.
9 with the values of the basic case. It is clear that the addition
of the biogeochemical changes have improved the overall
goodness of fit of the solution, especially near the surface
where the large differences between the data and the model
estimates have been significantly reduced by the inclusion of
the biogeochemical terms.
5 Conclusions
The OMP method is applied to a multi-year dataset of water
properties measured in the Straits of Juan de Fuca and
Georgia. The analysis is carried out on a seasonal basis with
six tracers (temperature, salinity, nitrate, phosphate, silicate,
and dissolved oxygen) and three main pre-defined source
water types. However, in winter, the analysis indicates that a
distinct fourth water source is present at the surface of Juan de
Fuca Strait near its mouth. It corresponds to winter intrusions
of relatively fresh warm water originating from the Columbia
River plume.
The analysis correctly identifies dense shelf water entering
at depth in Juan de Fuca Strait and mixing in the area of
strong tidal currents in Haro Strait with the fresh surface
plume of the Fraser River. The resulting mixed water then
advects out of Juan de Fuca Strait as a surface outflow and
into the Strait of Georgia as clearly identified subsurface
intrusions penetrating the Strait of Georgia at sill water depth
(around 100 m). In the deep Strait of Georgia, below sill
depth, water appears stagnant in both winter and early spring
but is displaced by deep water intrusions in summer and fall.
When a fifth source, representing the subsurface flow originating at the northern end of the Strait of Georgia, is added,
model results are consistent with the presence of subsurface
intrusions advecting southward from the northern channels at
a depth of about 80 m. Given the significantly smaller crosssectional areas of the northern channels, these intrusions are
of much lesser extent than the ones originating from the
south. However, although the addition of this water source
does improve the local performance of the model, it cannot
unequivocally confirm the presence of northern intrusions.
Finally, the OMP analysis is extended to include biogeochemical changes due to primary production and remineralization. A large decrease in nutrient concentrations and a
concomitant increase in dissolved oxygen in the freshwater
plume at the surface of the Strait of Georgia indicate a high
level of photosynthetic activity in the euphotic zone. This
process is at a maximum in April during the intense spring
bloom. On the other hand, the mixing model indicates a
Seasonal Water Mass Analysis for the Straits of Juan de Fuca and Georgia / 15
marked increase in nutrient and oxygen concentrations in the
deep Strait of Georgia associated with remineralization.
Acknowledgements
Comments from Sophia Johannessen and Patrick Cummins
on an early draft of the manuscript are appreciated. Thanks go
to Lizette Beauchemin for the preparation of the figures.
Some figures were prepared with the help of the Ocean Data
View software (Schlitzer, 2002). Thanks also go to the various crew members of the CCGS Vector.
References
CONWAY, K.W.; J.V. BARRIE
and M. KRAUTTER. 2004. Modern siliceous sponge
reefs in a turbid, siliciclastic setting: Fraser River delta, British Columbia,
Canada. Neues Jahrbuch für Geologie und Paläontologie, Monatshefte. 6:
335–350.
HARRISON, P.J.; P.J. CLIFFORD, W.P. COCHLAN, K. YIN, M.A. ST-JOHN, P.A. THOMPSON, M.J. SIBBALD and L.J. ALBRIGHT. 1991. Nutrient and plankton dynamics in the Fraser River plume, Strait of Georgia, British Columbia. Mar.
Ecol. Prog. Ser. 70: 291–304.
HERLINVEAUX, R.H. and J.P. TULLY. 1961. Some oceanographic features of
Juan de Fuca Strait. J. Fish. Res. Board Can. 18 (6): 1027–1071.
HICKEY, B.M. and N.S. BANAS. 2003. Oceanography of the U.S. Pacific
Northwest coastal ocean and estuaries with application to coastal ecology.
Estuaries, 26 (4B): 1010–1031.
HUPE, A. and J. KARSTENSEN. 2000. Redfield stoichiometry in Arabian Sea
subsurface waters. Global Biogeochem. Cycles, 14 (1): 357–372.
KARSTENSEN, J. and M. TOMCZAK. 1998. Age determination of mixed water
masses using CFC and oxygen data. J. Geophys. Res. 103(18):
18599–18610.
LASDON, L.S.; A.D. WAREN, A. JAIN and M. RATNER. 1978. Design and testing
of a generalized reduced gradient code for nonlinear programming. ACM
Trans. Math. Software, 4 (1): 34–50.
LEBLOND, P.H. 1983. The Strait of Georgia: functional anatomy of a coastal
sea. Can. J. Fish. Aquat. Sci. 40: 1033–1063.
——; H. MA, F. DOHERTY and S. POND. 1991. Deep and intermediate water
replacement in the Strait of Georgia. ATMOSPHERE-OCEAN, 29: 288–312.
LEWIS, A.G. 1978. Concentrations of nutrients and chlorophyll on a crosschannel transect in Juan de Fuca Strait, British Columbia. J. Fish. Res.
Board Can. 35: 305–314.
LI, M.; A. GARGETT and K. DENMAN. 1999. Seasonal and interannual variability of estuarine circulation in a box model of the Strait of Georgia and
Juan de Fuca Strait. ATMOSPHERE-OCEAN, 37 (1): 1–19.
MAAMAATUAIAHUTAPU, K.; V.C. GARCON, C. PROVOST, M. BOULAHDID and A. A.
BIANCHI. 1994. Spring and winter mass composition in the BrazilMalvinas Confluence. J. Mar. Res. 52: 397–426.
MACKAS, D.L.; K.L. DENMAN
and A.F. BENNETT. 1987. Least squares multiple
tracer analysis of water mass composition. J. Geophys. Res. 92(C3):
2907–2918.
MASSON, D. 2002. Deep water renewal in the Strait of Georgia. Estuar.
Coastal Shelf Sci. 54: 115–126.
—— and P.F. CUMMINS. 2004. Observations and modelling off seasonal variability in the Straits of Georgia and Juan de Fuca. J. Mar. Res. 62(4):
491–516.
PARSONS, T.R.; J. STRONACH, G.A. BORSTAD, G. LOUTTIT and R.I. PERRY. 1981.
Biological fronts in the Strait of Georgia, British Columbia, and their relation to recent measurements of primary productivity. Mar. Ecol. Prog.
Ser. 6: 237–242.
REDFIELD, A.C.; B.H. KETCHUM and F.A. RICHARDS. 1963. The influence of
organism on the composition of sea water. In: The Sea, Vol. 2, M.N. Hill
(Ed.), Wiley-Interscience, New York, pp. 26–77.
Schlitzer, R. 2002. Ocean Data View, http://www.awi-bremerhaven.de/
GEO/ODV.
Thomson, R.E. 1981. Oceanography of the British Columbia coast. Can.
Spec. Publ. Fish. Aquat. Sci., 56, 291pp.
TOMCZAK, M. and D.G.B. LARGE. 1989. Optimum multiparameter analysis in
the thermocline of the eastern Indian Ocean. J. Geophys. Res. 94 (C11):
16141–16149.
TULLY, J. P. and A.J. DODIMEAD. 1957. Properties of the water in the Strait of
Georgia, British Columbia, and influencing factors. J. Fish. Res. Board
Can. 14 (3): 241–319.
WALDICHUK, M. 1957. Physical oceanography of the Strait of Georgia, British
Columbia. J. Fish. Res. Board Can. 14 (3): 321–486.
YOU, Y. 1997. Seasonal variations of thermocline circulation and ventilation
in the Indian Ocean. J. Geophys. Res. 102 (C5): 10391–10422.
—— and M. TOMCZAK. 1993. Thermocline circulation and ventilation in the
Indian Ocean derived from water mass analysis. Deep-Sea Res. 40 (1):
13–56.